Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Use when user mentions fantasy, magic system, or world-building for fantastical settings - provides fantasy genre conventions, magic system design patterns, and world-building frameworks
.claude/skills/microck-fantasy-world-building/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-01 | ✗→✓ | ▲ Improved | 82% | 0% |
| case-02 | ✗→✓ | ▲ Improved | 35% | 0% |
| case-03 | ✗→✓ | ▲ Improved | 53% | 0% |
| case-07 | ✗→✓ | ▲ Improved | 81% | 0% |
| case-14 | ✓→✗ | ▼ Worse | 104% | 0% |
| 元素 | 指导原则 | 关键点 | |------|---------|-------| | 魔法系统 | 必须有清晰规则 | 限制比力量更重要 | | 世界设定 | 内在自洽 | 每个规则都有原因 | | 种族/生物 | 独特且有逻辑 | 避免单纯的人类翻版 | | 历史深度 | 至少三代历史 | 过去影响现在 | | 政治结构 | 权力分布清晰 | 冲突有根源 |
第一法则：读者从魔法中获得的满足感，与其理解魔法的程度成正比
第二法则：限制比力量更有趣
第三法则：在添加新东西之前，先扩展已有的
必要元素：
示例框架：
魔法系统：元素操控
来源：每个人出生时拥有一种元素亲和
规则：只能操控自己的元素，需要该元素存在于周围
限制：精神疲劳，过度使用导致元素反噬
代价：使用魔法时消耗生命力，需要休息恢复
禁忌：不能创造元素，只能操控；不能操控生物体内的元素特征：
使用场景：
必须考虑：
层次架构：
政治层面：
经济层面：
社会层面：
文化层面：
最少三代历史：
当代（故事发生时）：
父辈代（30-50年前）：
祖辈代（60-100年前）：
古代（更久远）：
避免单一特征： ❌ 所有精灵都优雅高贵 ❌ 所有矮人都贪婪暴躁 ❌ 所有兽人都野蛮好战
创造深度： ✅ 内部多样性（不同文化、价值观） ✅ 个体差异（性格各异） ✅ 历史复杂性（好的和坏的历史）
种族特征应该有原因：
设计原则：
问题：花费章节解释世界设定，停止故事推进
解决：
问题：魔法/世界规则为了情节便利而改变
解决：
问题：所有奇幻都是欧洲中世纪的翻版
解决：
问题：主角因为预言/血统特殊，而非行动
解决：
/specify 执行时/plan 期间/write 时/analyze 期间创造的 vs 展示的：
第一章：
前 25%：
中段：
后段：
奇幻读者想要什么：
让奇幻读者沮丧的是什么：
记住：伟大的世界构建是故事的基础，而非目的。世界应该服务于角色和情节，而角色的行动应该受到世界规则的塑造和限制。平衡深度与叙事流畅性是关键。
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-04 | pass→pass | 27,366 | 29,159 | +7% | 1 | 1 | 0% | 3,511 | 6,030 | +72% | 0 | 0 | — |
case-01 | fail→pass | 29,204 | 35,174 | +20% | 1 | 1 | 0% | 3,860 | 7,017 | +82% | 0 | 0 | — |
case-02 | fail→pass | 35,537 | 32,104 | -10% | 1 | 1 | 0% | 4,915 | 6,644 | +35% | 0 | 0 | — |
case-03 | fail→pass | 34,805 | 33,990 | -2% | 1 | 1 | 0% | 4,614 | 7,069 | +53% | 0 | 0 | — |
case-05 | fail→fail | 23,273 | 25,881 | +11% | 1 | 1 | 0% | 2,994 | 5,764 | +93% | 0 | 0 | — |
case-06 | pass→pass | 21,610 | 24,552 | +14% | 1 | 1 | 0% | 2,979 | 5,660 | +90% | 0 | 0 | — |
case-07 | fail→pass | 25,810 | 28,557 | +11% | 1 | 1 | 0% | 3,291 | 5,952 | +81% | 0 | 0 | — |
case-08 | fail→fail | 21,527 | 20,686 | -4% | 1 | 1 | 0% | 2,914 | 5,235 | +80% | 0 | 0 | — |
case-09 | pass→pass | 23,132 | 22,707 | -2% | 1 | 1 | 0% | 3,042 | 5,413 | +78% | 0 | 0 | — |
case-10 | pass→pass | 20,617 | 26,695 | +29% | 1 | 1 | 0% | 2,800 | 5,827 | +108% | 0 | 0 | — |
case-11 | pass→pass | 27,368 | 27,607 | +1% | 1 | 1 | 0% | 3,666 | 5,979 | +63% | 0 | 0 | — |
case-12 | pass→pass | 23,592 | 26,999 | +14% | 1 | 1 | 0% | 3,169 | 5,938 | +87% | 0 | 0 | — |
case-13 | fail→fail | 23,739 | 23,691 | -0% | 1 | 1 | 0% | 3,395 | 5,864 | +73% | 0 | 0 | — |
case-14 | pass→fail | 17,444 | 18,426 | +6% | 1 | 1 | 0% | 2,452 | 5,001 | +104% | 0 | 0 | — |
case-15 | pass→pass | 21,732 | 23,226 | +7% | 1 | 1 | 0% | 2,888 | 5,632 | +95% | 0 | 0 | — |
case-16 | pass→pass | 25,009 | 34,811 | +39% | 1 | 1 | 0% | 3,266 | 6,753 | +107% | 0 | 0 | — |
case-17 | pass→pass | 23,002 | 23,053 | +0% | 1 | 1 | 0% | 2,833 | 5,259 | +86% | 0 | 0 | — |
case-18 | pass→pass | 27,200 | 23,794 | -13% | 1 | 1 | 0% | 3,487 | 5,525 | +58% | 0 | 0 | — |
case-19 | pass→pass | 20,213 | 17,206 | -15% | 1 | 1 | 0% | 2,778 | 4,753 | +71% | 0 | 0 | — |
case-20 | pass→pass | 32,387 | 32,615 | +1% | 1 | 1 | 0% | 4,148 | 6,842 | +65% | 0 | 0 | — |
case-21 | pass→pass | 18,134 | 19,040 | +5% | 1 | 1 | 0% | 2,696 | 5,067 | +88% | 0 | 0 | — |
case-22 | pass→pass | 31,266 | 29,160 | -7% | 1 | 1 | 0% | 4,080 | 6,101 | +50% | 0 | 0 | — |
DecimalAI ran this skill against gemini-3.6-flash twice over the same eval suite — once with the skill loaded and once without — and compared the two runs case by case. 22 cases were attempted. The headline lift of +14 percentage points is the difference between those two pass rates over the 22 comparable cases. 1 case got worse with the skill loaded, and it is included in that figure.
Without the skill loaded, the model failed this case. With it loaded, the same prompt on the same model passed. This is one improved case from the latest verified run; every case, including any that regressed, is in the table above.
Other measured skills in the registry, with their headline benchmark lift.